Volatility Measures for Regime Classification

Regime detection and classification are important in portfolio management and asset allocation. One of the key inputs into regime detection models is volatility. Reference [1] examines which volatility measure is most effective for regime classification. The authors study three volatility measures,

1. Implied volatility (VIX/VIXBR),

2. GARCH conditional volatility,

3. Historical volatility.

They then incorporate them into a Hidden Markov Model framework. The paper pointed out,

The empirical evidence supports conditional volatility (GARCH as the superior proxy for regime identification in both the Brazilian and U.S. markets). Unlike implied volatility, which exhibited excessive sensitivity to short-term noise and threshold variations, the GARCH- based specification provided the greatest parameter stability and classification robustness, essential attributes for operationalizing dynamic portfolios.

A key finding of this research is the structural identification of three volatility regimes (low, medium, and high). Contrary to binary specifications often assumed in the literature for interpretability, the Bayesian Information Criterion (BIC) results demonstrated that a three-state model better captures the complex dynamics of financial markets, specifically identifying transitional phases that binary models fail to detect. This granular identification allowed for a more precise assessment of international risk transmission…

From an investment perspective, the results highlight the distinct roles of regime-based strategies. The Dynamic Regime strategy consistently outperformed the traditional Static Mean-Variance (Single Regime) strategy in risk-adjusted metrics, successfully mitigating severe drawdowns, most notably in the U.S. market, where the static strategy suffered structural losses (−18.49%) while the dynamic strategy preserved capital (+1.56%). However, the dynamic strategy did not outperform the Naive (1/N) benchmark in terms of total cumulative return…

In short, the authors conclude that GARCH conditional volatility provides the most stable and operationally reliable regime classification. Implied volatility reacts faster to market changes but produces noisier regime switching. The study also finds that a three-regime framework is superior to a simple low/high volatility classification, as the intermediate regime captures transition periods and uncertainty normalization phases.

Another important point emphasized in the paper is that regime-based strategies are best viewed as risk-management tools rather than universal return-enhancing solutions. Regime-based allocation improves drawdown control and risk-adjusted performance relative to static mean-variance optimization.

Let us know what you think in the comments below.

References

[1] Bitencourt, W. A., & Iquiapaza, R. A. (2026), Comparative analysis of volatility proxies and regime-based asset allocation, International Review of Economics and Finance, 109, 105366.